Great work !
Woohoo!
Thanks!
#
Dear TJ and Zak and and Everyone,
I am reposting Zaks’ answer for completeness, thanks Zak!
Many thanks to everyone who posted in this thread to raise awareness about this issue!
And special thanks to for helping us to find the solution!
#Quoting Zak
I just submitted a code fix, CL
2378178.
There is network smoothing code in CharacterMovement that sets the relative location of the mesh each update. I think this broke when the mesh is detached (when simulating physics), so that relative location now becomes a global location, usually close to the world origin.
A simpler workaround than the proposed one (though that helped me narrow it down, so thanks for that!) is to just SetComponentTickEnabled(false) on CharacterMovement before the ragdoll (this disables the code in question). Or if you have C++ code you can override UCharacterMovementComponent::SmoothClientPosition() with my new code, which basically does nothing if the mesh is simulating physics.
-Zak
Thanks TJ!
I am crediting above, but thanks for your help too! And you too Zak!